Sep 15 → Nov 5 · ~2 hrs/workday, 3–4 hrs/weekend day · Ready to apply by November
Priority order: DSA → System Design (+CS fundamentals) → ML/DL → MLOps. Each week is a page below; tick problems as you go. Boxes reset on reload (this is a checklist, not saved state) — track real progress in the tracker/LeetCode.
Daily routine (~2 hrs): 75 min DSA (NeetCode, timed) + 10 min syntax typing (SyntaxCache/SpeedCoder) + 35 min second block. Weekends: long DSA set + project + 1 system-design practice out loud.
Golden rules: talk out loud while solving · no AI autocomplete · after a working solution spend 60s on complexity (your timeout-ghost fix) · any problem over the time cap → redo from scratch 48h later.
Project 1 — Recommender System (Week 3 kickoff → Week 4 finish): two-tower/embedding recommender, Recall@K + NDCG, FastAPI-served, Docker. Fills the recsys gap.
Project 2 — Multi-model inference server + benchmark (Week 5 kickoff → Week 6 finish): serve 2–3 models, latency/throughput benchmark, monitoring, Docker + GitHub Actions CI. Public version of your Triton/GPU-scheduling story.
Only list what is actually on your public GitHub and defensible end to end.